项目名称: 高维协变量下部分线性风险回归模型的变量选择
项目编号: No.11201349
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 数理科学和化学
项目作者: 陈玉蓉
作者单位: 武汉大学
项目金额: 22万元
中文摘要: 生存分析研究中,考察协变量对失效时间的各种效应是研究的热点之一,其应用已渗透到各个领域。在现有生命科学研究中,经常会遇到生存数据是删失的,同时协变量又为高维的,远远大于研究个体数量的情形。高维协变量的产生使得经典生存分析方法不再适用。虽完全数据模型中高维数据处理方法日趋成熟,但应用于观测数据出现删失的风险模型相对较少,且主要集中在假设协变量对失效时间或危险率函数的线性效应以及假设生存数据独立的模型中,但非线性效应、加性效应在实际应用中大量存在,另外生存数据本身还可能是相关的,因此迫切需要研究这些情形下高维协变量的变量选择问题。本项目拟充分利用和发展变量选择理论,参数及半参数理论,数值计算理论等解决若干高维协变量下部分线性风险回归模型的变量选择问题,具体研究部分线性可加Cox模型等三个模型,并将研究成果应用到乳房瘤等实际数据中,进一步拓宽变量选择的理论研究思路和应用领域。
中文关键词: 风险回归模型;变量选择;高维协变量;部分线性;
英文摘要: In survival analysis, the investigation about the various effects of covariates on the failure time is one of the hot research topics. Its application has penetrated into many fields. Censored data are often met in survival science research while the number of covariates which are high-dimensional data may be far greater than the number of individuals. The classic survival analysis methods no longer apply under such circumstances. There is a few works in variable selection in the censoring models although variable selection methods for high-dimensional data in full data model has gradually come to maturity. These works foucs on the models which assume that covariates make linear effects and survival data is independent.However non-linear effcts ,additive effects or the survival be relevant are ofen met in practice.There is an urgent need to study the variable selection for high-dimensional covariates in these circumstances. The project is planned to make full use of and development of variable selection theory, the parametric and semi-parametric theory, numerical computation theory to solve variable selection with high-dimensional covariates in partially linear hazard regression models.Three models such as partially linear additive Cox model will be investigated. The study will supplement the variable selecti
英文关键词: hazard regression model;varibale selection;high-dimensional covariates;partially linear regression model;